/* BLIS An object-based framework for developing high-performance BLAS-like libraries. Copyright (C) 2014, The University of Texas at Austin Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: - Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. - Neither the name(s) of the copyright holder(s) nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "blis.h" #undef GENTFUNCCO #define GENTFUNCCO( ctype, ctype_r, ch, chr, opname, arch, suf, trsmkerid ) \ \ void PASTEMAC3(ch,opname,arch,suf) \ ( \ dim_t m, \ dim_t n, \ dim_t k, \ const void* alpha0, \ const void* a1x0, \ const void* a110, \ const void* bx10, \ void* b110, \ void* c110, inc_t rs_c, inc_t cs_c, \ auxinfo_t* data, \ const cntx_t* cntx \ ) \ { \ const ctype* alpha = alpha0; \ const ctype* a1x = a1x0; \ const ctype* a11 = a110; \ const ctype* bx1 = bx10; \ ctype* b11 = b110; \ ctype* c11 = c110; \ \ const num_t dt = PASTEMAC(ch,type); \ const num_t dt_r = PASTEMAC(chr,type); \ \ gemm_ukr_ft rgemm_ukr = bli_cntx_get_ukr_dt( dt_r, BLIS_GEMM_UKR, cntx ); \ \ trsm_ukr_ft ctrsm_vukr = bli_cntx_get_l3_vir_ukr_dt( dt, trsmkerid, cntx ); \ \ const bool col_pref_r = bli_cntx_ukr_prefers_cols_dt( dt_r, BLIS_GEMM_UKR, cntx ); \ \ const dim_t mr = bli_cntx_get_blksz_def_dt( dt, BLIS_MR, cntx ); \ const dim_t nr = bli_cntx_get_blksz_def_dt( dt, BLIS_NR, cntx ); \ \ const dim_t mr_r = bli_cntx_get_blksz_def_dt( dt_r, BLIS_MR, cntx ); \ const dim_t nr_r = bli_cntx_get_blksz_def_dt( dt_r, BLIS_NR, cntx ); \ \ ctype bt[ BLIS_STACK_BUF_MAX_SIZE \ / sizeof( ctype ) ] \ __attribute__((aligned(BLIS_STACK_BUF_ALIGN_SIZE))); \ inc_t rs_bt; \ inc_t cs_bt; \ \ inc_t rs_bt_r; \ inc_t cs_bt_r; \ \ const dim_t packnr = bli_cntx_get_blksz_max_dt( dt, BLIS_NR, cntx ); \ \ const pack_t schema_b = bli_auxinfo_schema_b( data ); \ \ const dim_t k2 = 2 * k; \ \ const ctype_r* a1x_r = ( ctype_r* )a1x; \ \ ctype_r* bx1_r = ( ctype_r* )bx1; \ \ const inc_t rs_b = packnr; \ const inc_t cs_b = bli_cntx_get_blksz_def_dt( dt_r, BLIS_BBN, cntx ); \ \ const ctype_r* zero_r = PASTEMAC(chr,0); \ const ctype_r* minus_one_r = PASTEMAC(chr,m1); \ \ const ctype_r alpha_r = PASTEMAC(ch,real)( *alpha ); \ const ctype_r alpha_i = PASTEMAC(ch,imag)( *alpha ); \ \ ctype_r* b_use; \ \ inc_t rs_b_use; \ inc_t cs_b_use; \ \ ctype ct[ BLIS_STACK_BUF_MAX_SIZE \ / sizeof( ctype ) ] \ __attribute__((aligned(BLIS_STACK_BUF_ALIGN_SIZE))); \ /* FGVZ: Should we be querying the preference of BLIS_GEMMTRSM_?_UKR instead? */ \ const bool col_pref = bli_cntx_ukr_prefers_cols_dt( dt, BLIS_GEMM_VIR_UKR, cntx ); \ const inc_t rs_ct = ( col_pref ? 1 : nr ); \ const inc_t cs_ct = ( col_pref ? mr : 1 ); \ \ const bool use_ct = ( m < mr || n < nr ); \ \ ctype* c11_use = c11; \ \ inc_t rs_c_use = rs_c; \ inc_t cs_c_use = cs_c; \ \ if ( use_ct ) \ { \ c11_use = ct; \ rs_c_use = rs_ct; \ cs_c_use = cs_ct; \ } \ \ \ /* Handle alphas with non-zero imaginary components. */ \ /* NOTE: This branch should never execute because alphas with non-zero imaginary components should be applied during packing, and so the only alphas we should see here are those exclusively in the real domain, either because the value originally had no imaginary compoent (e.g. 4.0) or because a 1.0 was sent in as a placeholder since the alpha was applied during packing. */ \ if ( 0 ) \ if ( !PASTEMAC(chr,eq0)( alpha_i ) ) \ { \ bli_abort(); \ \ /* ctype_r* restrict one_r = PASTEMAC(chr,1); \ \ const inc_t ld_b = rs_b; \ \ PASTEMAC(ch,scal1ms_mxn)( schema_b, \ mr, \ nr, \ alpha, \ b11, rs_b, cs_b, ld_b ); \ \ alpha_r = *one_r; \ */ \ } \ \ \ { \ /* Set the strides for the temporary bt matrix based on the native real domain micro-kernel storage preferences. */ \ if ( col_pref_r ) { rs_bt = 1; cs_bt = mr; \ rs_bt_r = 1; cs_bt_r = mr_r; } \ else { rs_bt = nr; cs_bt = 1; \ rs_bt_r = nr_r; cs_bt_r = 1; } \ \ b_use = ( ctype_r* )bt; \ rs_b_use = rs_bt_r; \ cs_b_use = cs_bt_r; \ } \ \ \ /* Since b11 is stored in the 1e or 1r schema, we cannot update it directly, and instead must compute the matrix product in a local temporary microtile and then accumulate it into b11 according to its schema. */ \ \ \ /* lower: bt = -1.0 * a10 * b01; upper: bt = -1.0 * a12 * b21; */ \ rgemm_ukr \ ( \ mr_r, \ nr_r, \ k2, \ minus_one_r, \ a1x_r, \ bx1_r, \ zero_r, \ b_use, rs_b_use, cs_b_use, \ data, \ cntx \ ); \ \ \ if ( bli_is_1e_packed( schema_b ) ) \ { \ const inc_t ld_b = rs_b; \ const inc_t rs_b2 = 2 * rs_b; \ const inc_t cs_b2 = 2 * cs_b; \ \ ctype_r* restrict b11_ri = ( ctype_r* )b11; \ ctype_r* restrict b11_ir = ( ctype_r* )b11 + ld_b; \ \ /* b11 = alpha * b11 + bt; */ \ for ( dim_t j = 0; j < nr; ++j ) \ for ( dim_t i = 0; i < mr; ++i ) \ for ( dim_t d = 0; d < cs_b; ++d ) \ { \ ctype* restrict beta11t = bt + i*rs_bt + j*cs_bt; \ ctype_r* restrict beta11t_r = &PASTEMAC(ch,real)( *beta11t ); \ ctype_r* restrict beta11t_i = &PASTEMAC(ch,imag)( *beta11t ); \ \ ctype_r* restrict beta11_ri_r = b11_ri + i*rs_b2 + j*cs_b2 + 0*cs_b + d; \ ctype_r* restrict beta11_ri_i = b11_ri + i*rs_b2 + j*cs_b2 + 1*cs_b + d; \ ctype_r* restrict beta11_ir_r = b11_ir + i*rs_b2 + j*cs_b2 + 0*cs_b + d; \ ctype_r* restrict beta11_ir_i = b11_ir + i*rs_b2 + j*cs_b2 + 1*cs_b + d; \ \ PASTEMAC3(ch,chr,ch,xpbyris) \ ( \ *beta11t_r, \ *beta11t_i, \ alpha_r, \ alpha_i, /* alpha_i not referenced */ \ *beta11_ri_r, \ *beta11_ri_i \ ); \ \ PASTEMAC(ch,copyris)( -*beta11_ri_i, *beta11_ri_r, \ *beta11_ir_r, *beta11_ir_i ); \ } \ } \ else /* if ( bli_is_1r_packed( schema_b ) ) */ \ { \ const inc_t ld_b = rs_b; \ const inc_t rs_b2 = 2 * rs_b; \ const inc_t cs_b2 = cs_b; \ \ ctype_r* restrict b11_r = ( ctype_r* )b11; \ ctype_r* restrict b11_i = ( ctype_r* )b11 + ld_b; \ \ /* b11 = alpha * b11 + bt; */ \ for ( dim_t j = 0; j < nr; ++j ) \ for ( dim_t i = 0; i < mr; ++i ) \ for ( dim_t d = 0; d < cs_b; ++d ) \ { \ ctype* restrict beta11t = bt + i*rs_bt + j*cs_bt; \ ctype_r* restrict beta11t_r = &PASTEMAC(ch,real)( *beta11t ); \ ctype_r* restrict beta11t_i = &PASTEMAC(ch,imag)( *beta11t ); \ \ ctype_r* restrict beta11_r = b11_r + i*rs_b2 + j*cs_b2 + d; \ ctype_r* restrict beta11_i = b11_i + i*rs_b2 + j*cs_b2 + d; \ \ PASTEMAC3(ch,chr,ch,xpbyris) \ ( \ *beta11t_r, \ *beta11t_i, \ alpha_r, \ alpha_i, /* alpha_i not referenced */ \ *beta11_r, \ *beta11_i \ ); \ } \ } \ \ \ /* b11 = inv(a11) * b11; c11 = b11; */ \ ctrsm_vukr \ ( \ a11, \ b11, \ c11_use, rs_c_use, cs_c_use, \ data, \ cntx \ ); \ \ if ( use_ct ) \ { \ PASTEMAC(ch,copys_mxn) \ ( \ m, n, \ ct, rs_ct, cs_ct, \ c11, rs_c, cs_c \ ); \ } \ } INSERT_GENTFUNCCO( gemmtrsm1m_l, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, BLIS_TRSM_L_UKR ) INSERT_GENTFUNCCO( gemmtrsm1m_u, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, BLIS_TRSM_U_UKR )